HKUDS

Simple, fast, and strong LLM agent framework with tool use, computer use, and orchestration.

Social AI agent that uses OpenClaw to handle your DMs, posts, and online interactions on your behalf.

Vibe writing assistant for faster, better drafting — write with AI in a lightweight editor.

AI research assistant that sparks novel ideas and accelerates scientific discovery — point it at a domain and get hypothesis suggestions.

Personal AI agent that learns from your screen and recall history to deliver context-aware responses.

WSDM 2024 paper using LLMs with graph augmentation for recommendation systems, including multi-modal and side-information signals.

Self-supervised learning framework for recommender systems. Covers contrastive, generative, and graph SSL methods in one toolkit.

ICML 2025 paper accelerating large language models by compressing each segment into a single separator token to speed up inference.

Universal tool-use layer for AI agents, providing orchestration over arbitrary tools and MCP servers for computer-use agents.

ACL 2026 paper exploring whether small language models can teach large language models to reason, with code and training recipes.

All-in-one agentic framework for video understanding, editing, and remaking. Open-source counterpart to NotebookLM-style video tools.

HKUDS's RAG system for video — chat with hours of footage, ask questions, and get cited timestamps and clips back.

HKUDS's system for accelerating and streamlining code understanding — LLM-powered repo navigation, symbol mapping, and concept extraction at speed.

HKUDS's tool that turns a research paper into a presentation deck in one click — automatic slide generation with figures, key takeaways, and structure.

HKUDS's autonomous scientific innovation agent — proposes research questions, runs experiments, and writes up findings end-to-end.

Open agentic coding system from HKUDS — turns research papers, text descriptions, and design intent into working web frontends and backend services.

Fully-automated, zero-code LLM agent framework from HKUDS — define agents in natural language with no Python required.

Lightweight open-source AI agent that works across your tools, chats, and workflows — integrates with Claude, OpenAI Codex, and OpenClaw.

Simple, fast retrieval-augmented generation system from HKUDS that combines vector search and graph-based knowledge representation. EMNLP 2025.

A framework that automatically generates command-line interfaces for any software application, making them accessible to AI agents. Transforms GUI-based tools into agent-controllable CLIs with a unified interface across thousands of applications.

A self-evolving engine that makes AI agents smarter, more cost-efficient, and able to share knowledge with each other. Skills automatically improve from real-world usage, reducing token costs by 46% while enabling collective intelligence across agents.

🚀 RAG-Anything: All-in-One RAG Framework

Agent-native personalized learning assistant that combines unified chat workspace, autonomous TutorBots, AI co-writing, and guided learning paths. Features persistent memory and knowledge bases to create adaptive, context-aware tutoring experiences.